Global occlusion self-adaptive pedestrian training/recognition method, system and device and medium

A training method and pedestrian recognition technology, applied in the field of image recognition, can solve problems such as no intuitive application, achieve high industrial utilization value, reduce labor costs, and improve efficiency

Pending Publication Date: 2021-01-15
WINNER TECH CO INC
View PDF7 Cites 3 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0009] In view of the shortcomings of the prior art described above, the purpose of the present invention is to provide a global occlusion adaptive pedestrian training / recognition method, system and storage medium, which are used to solve the problem of pedestrians in different directions, different postures, and different cameras in the prior art. The occlusion phenomenon caused by the angle, and a new evaluation index is proposed to solve the problem that the existing evaluation index of the identification model is not directly reflected in the engineering application.

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Global occlusion self-adaptive pedestrian training/recognition method, system and device and medium
  • Global occlusion self-adaptive pedestrian training/recognition method, system and device and medium
  • Global occlusion self-adaptive pedestrian training/recognition method, system and device and medium

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0050] This embodiment provides a global occlusion adaptive pedestrian training method, including:

[0051] Receive a training data set; the training data set includes N pedestrians, each pedestrian has M pictures; wherein, N is greater than 1, and M is greater than 1;

[0052] Extract the attribute feature map of each picture to output the NxM attribute feature map;

[0053] Fuse all the attribute feature maps of the same pedestrian to form the fusion feature of the pedestrian, and obtain the fusion features of several pedestrians;

[0054] extracting local features from the plurality of attribute feature maps and extracting global features from the fusion features of the plurality of pedestrians;

[0055] According to the local feature and the global feature, the attention of the local feature is extracted to calculate the local feature attention-enhanced feature used to characterize the local feature and the global feature used to characterize the global feature. The feat...

Embodiment 2

[0133] This embodiment provides a global occlusion adaptive pedestrian training system, which is characterized in that it includes:

[0134] The data receiving module is used to receive a training data set; the training data set includes N pedestrians, and each pedestrian has M pictures; wherein, N is greater than 1, and M is greater than 1;

[0135] The first feature extraction module is used to extract the attribute feature map of each picture, so as to output the NxM attribute feature map;

[0136] The fusion module is used to fuse all the attribute feature maps of the same pedestrian to form the fusion feature of the pedestrian, and obtain the fusion features of several pedestrians;

[0137] The second feature extraction module is used to extract local features from the multiple attribute feature maps and extract global features from the fusion features of the plurality of pedestrians;

[0138]The attention extraction module is used to extract the attention of the local f...

specific Embodiment

[0154] Step 1: The data receiving module collects images of pedestrians, or downloads public datasets for pedestrian re-identification; divides the datasets into training sets and test sets; the Market1501 public datasets are used in this invention.

[0155] Step 2: The first feature extraction module is loaded into the VGG-16 network, and the initial weight is the pre-training weight of VGG-16 on ImageNet; for the second feature extraction module and the convolution layer in the attention extraction module, batch normalization The weights are initialized with a normal distribution with a mean value of 0 and a mean square error of 0.01, and the bias is initialized with 0. The alpha parameter value in TripletLoss is set to 0.3.

[0156] Step 3: Input data and train the network. Each batch of data includes 16 pedestrians, and each pedestrian has 4 images. The training is carried out for 100 epochs in total, the initial learning rate is set to 0.002, and the learning rate is mu...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

PUM

No PUM Login to view more

Abstract

The invention provides a global occlusion self-adaptive pedestrian training / recognition method, system and device and a medium. The pedestrian training method comprises the steps of receiving a training data set; extracting an attribute feature map of each picture; fusing all attribute feature maps of same pedestrians to obtain fused features of a plurality of pedestrians; extracting local features from the plurality of attribute feature maps and extracting global features from the fusion features of the plurality of pedestrians; extracting attention of the local features according to the local features and the global features so as to calculate features after attention enhancement of the local features used for representing the local features and features after shielding of the picture shielding information of the global features used for representing the global features; and performing loss calculation on the features after the attention of the local features is enhanced and the features after the picture shielding information of the global features is shielded. According to the invention, the model error caused by manual operation deviation is reduced; the problem that common model evaluation indexes cannot guide engineering threshold setting is solved.

Description

technical field [0001] The invention belongs to the technical field of image recognition, and relates to a training / recognition method and system, in particular to a global occlusion self-adaptive pedestrian training / recognition method, system and medium. Background technique [0002] Pedestrian re-identification refers to the technology of describing and comparing pedestrians in images or video sequences through computer vision methods, and judging whether there is a specified pedestrian. Pedestrian re-identification is a challenging subject due to the different environmental conditions, pedestrian poses, occlusions, and camera angles in different images or videos. [0003] Early pedestrian attribute recognition mainly constructed the apparent features of pedestrians by artificially selecting information such as the color and texture of pedestrian images, and trained classifiers in a targeted manner. Typical examples are HOG, SIFT, SVM, CRF models, etc. But these traditio...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

Application Information

Patent Timeline
no application Login to view more
Patent Type & Authority Applications(China)
IPC IPC(8): G06K9/00G06K9/62
CPCG06V40/103G06F18/253G06F18/214
Inventor 成西锋马卫民袁德胜游浩泉林治强党毅飞崔龙李伟超王海涛
Owner WINNER TECH CO INC
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap